[Ace-users] Re: [ciao-users] Component stoping with Dance
Gan Deng
gan.deng at vanderbilt.edu
Mon Sep 3 12:15:59 CDT 2007
Hi Steve,
James, Steven wrote:
> Hi Gan,
>
> Looking at the
> http://www.omg.org/news/meetings/workshops/RT_2006_Workshop_CD/071_Deng_
> eta.pdf
>
> I could use Add_instance and remove_instance?
I'm sorry that the above slides are a bit outdated. Surely you can add
and remove instances, but you must use the below operation on the
ExecutionManager interface to do that:
void perform_redeployment (in DeploymentPlan new_plan)
raises (PlanError,
InstallationFailure,
UnknownImplId,
ImplEntryPointNotFound,
InvalidConnection,
InvalidProperty,
::Components::RemoveFailure);
Assume you have an existing deployment plan, and you want to remove one
instance from it, then you can create a new plan which is the almost the
same as the existing deployment plan without the component in interest.
DAnCE provides a utility class (DAnCE_Utils) that can facilitate this.
For example, you can use below method in the utility class to manipulate
your deployment plan.
DAnCE_Utils::remove_instance (::Deployment::DeploymentPlan
&deployment_plan, const char *instance_name)
After this method, you can get a new deployment plan with the component
in interest removed, then you can call the perform_redeployment() in the
ExecutionMamanager to do the actual removal.
> Is ReDaC part of the OMG's D&C standard? And what version of CIAO is
> this part of I can't see it within 0.5.10?
No, ReDaC is CIAO specific. Since OMG's D&C is motivated by the JTRS
(joint tactical radio system), which is quite static, it does not
provide any mechanism to do the redeployment and reconfiguration.
CIAO 0.5.10 provides the ReDaC functionality, and you can follow the
step-by-step tutorial to work through an example.
> Furthermore, how redundant are the ReDaC and Dance demons i.e. what
> happens if a NodeManager dies for instance? Is this documented anywhere?
ReDaC does not require any additional daemons than the D&C spec-defined
ones, instead, ReDaC provides you the functionality by extending the
CCM-defined interfaces.
Regarding your question of NodeManager failure (which is typically
caused by node failure), it is an orthogonal issue, since it requires
special fault tolerant and fault recovery support. Andy, could you
please give Steve some input about this?
Thanks,
Gan
More information about the Ace-users
mailing list